Anthropic’s business depends on a demanding equation: keep attracting customers and improving its models while meeting enormous long-term commitments for compute and infrastructure. Its S-1 describes strong revenue growth and a focus on enterprise work, but also obligations that make continued expansion central to the company’s prospects.
Revenue growth meets a costly model
Anthropic’s revenue grew fourteen times to $16.3 billion between the first half of 2025 and the first half of 2026. The source article says the company reported an $80 billion loss, largely tied to an accounting matter it considers unrelated to the underlying business. After removing that and other charges, including employee compensation, adjusted income was $1.62 billion in the first half of 2026.
Those figures suggest that AI services can generate substantial income for Anthropic at present. But the company’s financial picture also depends on spending heavily to develop models and secure the computing capacity needed to serve customers. The reported results capture a period of rapid adoption; they do not by themselves establish that the same pace can continue.
A focused customer strategy
Anthropic has concentrated on coding and cybersecurity and has pursued enterprise customers. The article contrasts that approach with consumer-facing products such as image generators and TikTok-style services. Enterprise demand appears particularly important: about three-quarters of customers in the first half of 2026 were spending at least $100,000 a year, and net revenue retention was 900 percent.
This focus gives Anthropic a clear commercial case: tools that help businesses with computer-related work. Yet enterprise success does not remove competitive pressure. The source identifies OpenAI and Google as rivals, as well as open-source models that may appeal to users willing to accept software that is four months behind.
Anthropic’s position at the top of the Epoch Capabilities Index with Claude Opus 5.5 is also not a permanent advantage. The leaderboard changes frequently, and maintaining a leading model requires continued development. That creates a cycle in which new models can bring bursts of revenue and increased usage, while also demanding further investment.
Commitments stretch far into the future
The S-1 lists a series of obligations tied mainly to compute and infrastructure. It describes $265.6 billion in fixed minimum non-cancelable purchase commitments, with terms between one and 20 years. Lease agreements primarily related to compute and infrastructure capacity that had not yet commenced totaled $321.4 billion as of the unaudited June 30th figures.
Anthropic then entered additional compute and infrastructure agreements totaling about $245 billion, also with terms between one and 20 years. The filing further describes transactions that had not been finalized, estimated to cost $125 billion to $150 billion, along with variable compute commitments.
Adding the listed figures, the article estimates a minimum of around $850 billion in commitments over the next 20 years. Including the unfinished transactions at their current estimated cost would bring the total to about a trillion dollars. The author cautions that this calculation may omit items and notes that the S-1 does not provide a payment schedule for future commitments.
Growth is doing essential work
Analyst Harrison Rolfes of PitchBook models the commitments at about $90 billion a year from 2027 to 2030. On that basis, he estimates Anthropic would need $188 billion in revenue to keep its current ratio of cost to compute stable. Rolfes writes, “Growth is not optional for this company,” and adds, “The contracts assume it.”
Spending already shows the weight of development. Compute and infrastructure associated with customers cost $2.3 billion in the first half of this year, while model development cost $9.7 billion. The article says Anthropic has also become more efficient, spending less on compute to bring in each dollar of revenue, according to Rolfes.
That efficiency matters, but the company still faces a race against competitors with different advantages. Google can sell AI alongside products businesses already use, giving it a distribution advantage. Anthropic’s ability to keep growing will therefore depend on more than building capable models: it must turn those capabilities into lasting enterprise demand while managing long-term costs.
The S-1 presents both a business with striking momentum and a plan that leans heavily on future expansion. If revenue, usage, and efficiency keep improving, the commitments may support growth. If model gains or customer demand slow, the scale of those obligations makes the company’s dependence on continued progress harder to ignore.